Properties of Noise Cross Correlation Functions Obtained from a Distributed Acoustic Sensing (DAS) Array at Garner Valley, California

Abstract:

The field test of Distributed Acoustic Sensing (DAS) conducted at Garner Valley, California on September 11-12, 2013 provided a continuous overnight record of ambient noise. The DAS array recorded ground motions every one meter of optical cable that was arranged approximately in the shape of a rectangle with dimensions of 160 m by 80 m. The long dimension of the array was adjacent to a state highway. Three hours of record were used to compute noise cross-correlation functions (NCFs) in one-minute windows. The trace from each sensor channel was pre-processed by downsampling to 200 Hz, followed by normalization in the time-domain and bandpass filtering between 2 and 20 Hz (Bensen et al., 2007). The one-minute NCFs were then stacked using the time-frequency domain phase-weighted stacking method (Schimmel & Gallart, 2007). The NCFs between channels were asymmetrical reflecting the direction of traffic noise. The group velocities were found using the frequency-time analysis method. The energy was concentrated between 5 and 15 Hz, which falls into the typical traffic noise frequency band. The resulting velocities were between 100 and 300 m/s for frequencies between 10 and 20 Hz, which are in the same range as described in the results for surface-wave dispersion obtained using an active source for the same site (Lancelle et al., 2015). The group velocity starts to decrease for frequencies greater than ~10 Hz, which was expected on the basis of a previous shear-wave velocity model (Steidl et al., 1996). Then, the phase velocity was calculated using the multichannel analysis of surface wave technique (MASW - Park et al., 1999) with 114 NCFs spaced one meter apart. The resulting dispersion curve between 5 and 15 Hz gave phase velocities that ranged from approximately 170 m/s at 15 Hz to 250 m/s at 5 Hz. These results are consistent with other results of active-source DAS and seismometer records obtained at the Garner Valley site (e.g., Stokoe et al. 2004). This analysis is part of the PoroTomo project (Poroelastic Tomography by Adjoint Inverse Modeling of Data from Seismology, Geodesy, and Hydrology, http://geoscience.wisc.edu/feigl/porotomo).